Stability and Generalization in Looped Transformers

arXiv:2604.15259 · cs.LG, cs.AI · Submitted 2026-04-16 · Read on arXiv

cs.LG, cs.AI

Submitted: 2026-04-16

Updated: 2026-11-04

Comments: 10 main pages, 30 total. Accepted to NeurIPS 2026 Main Track

Code: https://github.com/ashlab11/generalization

Project page: https://t-dillon.github.io/tdoku

License: http://creativecommons.org/licenses/by/4.0/

The gist: Looped transformers promise test-time compute scaling by spending more iterations on harder problems, but it remains unclear which architectural choices let them extrapolate to harder problems at…

Terminology

Abstract

Looped transformers promise test-time compute scaling by spending more iterations on harder problems, but it remains unclear which architectural choices let them extrapolate to harder problems at test time rather than memorize training-specific solutions. We introduce a fixed-point based framework for analyzing looped architectures along three axes of stability -- local attractivity, input-dependence, and geometry -- and use it to characterize when fixed-point iteration yields meaningful predictions. Theoretically, we prove that looped networks without recall have countable fixed points and cannot achieve strong input-dependence at any spectral regime, while recall with outer normalization reliably produces a regime in which fixed points are simultaneously locally-attractive, locally smooth in the input, and supported by stable backpropagation. Empirically, we train small looped transformers on chess, sudoku, and prefix-sums and find that downstream performance tracks the framework's predictions across tasks and architectural configurations. We additionally introduce internal recall, a novel recall placement variant allowing an identity residual path, and exploit its contrasting fixed-point behavior to test our framework's predictions in a controlled setting.

Sources

Related papers